change the dice generator.

[Romain672]

I doubt you worked in IT for 30 years, or if you did, I doubt you know well the subject of random generator.


Because if you have a code where you can predict the next number generated one time out of one billon, then if you use it to connect to the nuclear bomb terminal, that can be pretty bad.
If you use it to have 0.1% more chance to win in your next Catan game, nobody care. (if that happen once)

That's a childish comparaison, but it's to show why having something which work really well is enough for a game like Catan, but will not be enough for something which need security like bank account.

===

I strongly believe that humans can't dissociate a serie of random generated numbers truly random from the random fonction used by bga. And if they can't, then that's enough good for hobbies uses.

[Romain672]

Note that here you give interesting claims.

How is it the Spies Cards always come out together? If you get a spy, you use it on the guy with the other blue card to get the other spy from them.
How is it that the Inventor Cards come out together? If you get one, and there are no blue cards because if there is, it will be the spy, and another person gets a green card you wait because the chances are the other inventor in in their hands.

I'm not understanding your claims here. That could be a langage barrier.

After more than 50% of a game goes by with out a number rolled to have that number roll 3 or 4 times in a row after the robber is moved there. This happens so often I am amazed no one talks about it.

Can you simplify that claim? That claim look the perfect example of humans having a biais. You see that you are in a 8. 8 isn't rolled you are mad. Someone put the robber there, so you are even more mad. Then 8 is rolled. And so ofc, that's because of the random.
Why would that be true and not 'the number where the robber is present is more rolled than the average'? And if that's the case too, then it would be easy to see that something is weird with the robber.
I could do stats on that claim, but... Do I really need to? That's a complicated claim, and that illustrate perfectly emotions even without doing any calculations.

[Theana]

I can understand Casinos. Sports Betting and Security systems using anything to take advantage of players..but what is the scope for BGA to do so?

[LukeDan]

Not that I think anything suspect is going on, randomness just happens and we notice patterns, but I just played in a game where 11 was rolled more often than 7.

[DivergentRalph]

Hi, I believe this topic needs investigation and a lot of data in order to come to a conclusion. Otherwise it will be people pointing out bizarre things and others saying it's just chance.

I've only played a couple games but have had a glimpse of some stuff happening that doesn't seem likely (though this could be confirmation bias, again, more data needed).

Last game:

2 3 4 5 6 7 8 9 10 11 12

4 1 3 6 4 9 18 3 3 2 3 (frequencies)

Using binomial distribution: X~B(56, 5/36) where 56 is number of rolls and 5/36 is probability rolling an 8

P(X greater than or equal to 18) = 0.00038899432 or 1 in 2571 games.

Very "lucky" that my second ever game is like this! Needless to say, the person next to the 8s won even with no access to sheep or brick.

In future I will document all the dice rolls of the games.

[Romain672]

I double checked, yes 1/2571 is right.
Now, you can do that claim for any number rolled too much or too few. That give you close to 22 claims possible, so let's multiply that probability per 20.
We got 1/129.

So yeah with only 2 games played that's weird, especially for you, but just play more, you will see that this game was extreme
That's far from the first subject of the type which arise.

An alternative would be that a recent change changed something, but there was nothing which changed recently it seem, and so we got no reason to think that is the case. (2 recent example start to be slightly weird, but that's still expected to happen over the years)
And still, if something recent is wrong, it's much more likely that we see a big change (like seeing probabilities of 1/1M or even weirder), than a minor change of probabilities.

[gmrsmrt]

Hi, I believe this topic needs investigation and a lot of data in order to come to a conclusion. Otherwise it will be people pointing out bizarre things and others saying it's just chance.

Right, so gather more data...

In the example you spotted you should add the probability of having any other number (2 to 12) to exceed the expected frequency by
a probability smaller or equal to the one you observed. And also the probability for each number to underscore the expected frequency by
a probability smaller or equal to the one you observed.
Once you do this you you get a probability 10 20 times the one you quote.
A relatively small number which occurs tenth or hundredth of times on BGA each day.

[Romain672]

Btw, the way I analyse the game let you just play all your games in your profile without collecting any data, that's the really nice upside

The thing is, people have a lot of difficulties of coming with games which are out of the ordinary from their whole profile. We got two very recent examples, but that's not what people usually share.

If something was wrong, I would expect to see from time to time profils with 1000+ games which find (after removing some biais) probabilities of less than 1%. The last big profile was still a 30%.

That really show that if something is wrong, it's not from rolling one specific number too much during a single game.
And that should make sense, it's something pretty easy to check and to be seen in games and by people, but yet we got lots of complains about that, included yours.

Outside of that, we regularly got lots of differents claims (drawing the same development cards in a row, rolling too much 2/3 number, rolling too much a number after half the game, rolling a number after the brigand just moved to it, and so on...).
There is some of them that I checked because the person was insistant or because there was multiple people telling that. And then comparing it to random numbers created from googlesheet (which could be bad because they could generate numbers the same way), but I find nothing really weird.

Then we got people giving no claim. And that's usually the most difficult kind of people to convince (at least enough for them to not post conspiracy on the forum). They will be looking at their game, find a new weird thing to complain about, and share it to the forum. But since their claim are various, they will find something weird every other game. But that tell nothing since there is 1000+ things they could check. And any calculation doesn't make much sense with those claim. I usually ask them to define a claim, and then share game about that claim in the future, but that never happen.

So yes... That will never end... People are biaiased, and are seeing pattern where there is nothing to see. You should look at that subject in the internet.
That's the same for many other websites too.

.

But if you are motivated, do it collect stats, few people do it, and you will get much more interesting answers about the statistical part.

[King Derf]

I could not agree more. T when you have events such as no 6 for 50 rolls and four 4s in a row which are statistically so highly unlikely it is ridiculous occur every game you have to ask what is going on? Having one in 200000 events occur every game is not right.

[Romain672]

I could not agree more. T when you have events such as no 6 for 50 rolls and four 4s in a row which are statistically so highly unlikely it is ridiculous occur every game you have to ask what is going on? Having one in 200000 events occur every game is not right.

No 6 in 50 rolls is 0.057%.
But, if your game has like 70 rolls total, then the chance to get one serie of 50 rolls is 1.2%.
If you add all numbers you goes to 12.4%.
If you add numbers rolled too much, you goes to 23.3%.

Three 4s in a row has a probability of 0,058% per turn. (i took three and not four because the first roll is free, that let you put all numbers together since you could complain about too much 5s too).
If you have 70 rolls, it's 3.8%.
And that's a number which work for the whole game. So 1 game out of 26, you should have a number rolled too much in a row than those four 4s.

So yeah, 23.3% and 3.8%.
So I'm sure you shared your extremes.


Btw your claims are extremely easy to convince me. In your future games, find me those weird things happenings. If really what you said was true, you will find those % every game. and you shouldn't get those 1/4 or 1/26 every game. (note that those probabilities should be again twice as unlikely since you talked about two differents things)